yejg2017 / practical3

Optimization and logistic regression

Practical 3

Machine Learning, spring 2015

Setup

Setup will be the same as last time in practical 1. Please refer to the practical 1 repository, and run the script as instructed last time.

We will use the Torch package gnuplot for plotting; you may use (advanced: recommended to try at home instead or if you have time afterwards) iTorch instead as the script installs it.

Materials

See simple_example.lua for the example mentioned in the "Introduction" part of the writeup, and practical3.lua for the template for the primary task.

We'll be classifying handwritten digits in a dataset called MNIST, which look like the following:

Each datapoint is a 32x32 image in the version of MNIST we have. The code provided will turn this into a vector containing all the raw pixel values. In simple_example.lua, there is a line containing the word "UNCOMMENT" illustrating how to view a digit from the train/test sets.

See course page for practicals

https://www.cs.ox.ac.uk/people/nando.defreitas/machinelearning/